Re: WHAT THE DEMO WILL DEMONSTRATE --voting methods, tabulation, and on-screen ballot

From: Alan Dechert <adechert_at_earthlink_dot_net>
Date: Tue Aug 26 2003 - 15:24:41 CDT

>
> I am willing to take the lead on the tabulation work with different voting
> methods, ...
>
That's great, Arnie. Thanks.

> ... but I am not a programmer, so I will need to work with someone.
>
We will get someone to work with you. Volunteers? I can help some too.

> I can take the data and create aggregated results for precincts and
> centralized districts and produce a colorful spreadsheet/table of results.
>
I see your deliverable as a CD with some software and some dummy data that
we can install on a laptop. This would be used for giving a presentation
about tabulation with our system. This is necessarily limited in scope -- a
level-of-effort project. I don't see this as a self-running
self-explanatory demo. You will have a menu of items listing things that we
can show. If we get lucky with equipment, we might show it with a
projector.

Time permitting, we could provide something that people could try on the
web. This might show just a few things.

One thing we need to resolve before you can get very far is the file format.
What will the uncompressed plain text ballot image look like? Do we have
116 columns with X indicating the vote and blank for no vote, with the
write-ins listed after that? It might be easier -- and more comprehensible
for the audience -- to display the selections for the 13 races comma
delimited, with any write-ins inserted into the appropriate column.

> Although we could do all of the races, it may be too much to do for a
demo.
>
I think we show,

1) Precinct level tabulation
2) County wide tabulation

Statewide and nationwide tabulation probably not. We can state that the
full study will include a computer simulation of the tabulation for a
general election with 100 million ballots.

At the outset, I think it's important to state what the demo will show and
what it will not show. We don't want to get questions that we answer, "Now
that's something we didn't think about." In such cases, we want to point to
the chart from the beginning and state "That's beyond the scope of the demo.
We understand your question and that will be addressed in the full study
where we will be working on the production system."

> Who are the audiences for the demo? Election officials and funding
> agencies--including NSF?
>
I think our primary target is the general public via the reporters that will
try out the system and report on it. Certainly, we will also target
election officials but keep in mind they know a lot about election
management software.

We should talk to elections people before getting too far along to make sure
we don't do things they will ridicule. And we want to get some idea of what
they'd like to see in such a demo. They'll probably want to see more than
we can give them but we can try to scale-down their expectations for the
demo -- "It's just a demo, folks!" Warren Slocum, elections chief of San
Mateo County would be a good person to start with. I've also met with John
Mott-Smith, the elections chief for the State of California SOS. He will
probably be willing to talk with us.

Funding agencies probably won't see the tabulation demo. But maybe it's
enough that we can say we did such a thing, and provide a brief description
or report.

> I agree that we should choose enough races to
> illustrate the tabulation of the results under different voting methods,
but
> I need more information about the reasons for defining the voting rules on
> your proposed ballot.
>
> For example, is voting for up to three candidates typical of California
> ballots? Is that the reason that you framed the cat catcher choice this
way?
> Unless it will rankle election officials in California, I would suggest
> allowing voters to cast one vote for each choice that they "approve." This
> framing of the vote will attract the attention of the national (and
> international) movement to advance approval voting. This movement is very
> concerned about technological impediments to the use of approval voting,
so
> our demonstration audience should be planned to invite state/local
> advocates.
>
I'm no expert on this. But I believe we use n-of-m voting to select county
party central committees. You get to pick several candidates to fill that
number of seats from a list of many candidates. The approval voting you
speak of, I believe, is designed to select a single winner. So, you may
want to demonstrate a couple of different scoring methods for cat catcher:
one where there are to be three cat catchers and another where there will be
one winner.

Doug?

> For the county commissioner race, the ranking of choices will appeal to
the
> advocates of Instant Referendum Vote at fairvote.com, ...
>
I think you mean fairvote.org

> ... particularly if we show the steps in taking an indecisive outcome
> and transforming it into a majoritarian outcome using the IRV methodology.
>
Yes.

> But IRV advocates seem
> oblivious to the strategic implications of collecting rank orderings. In
> fact, they point to a couple pieces of software that distort voter input
by
> preventing a voter from, say, ranking Packard and Hewlitt in first place
to
> show that there is a tie. Our interface should allow ties and make use of
> the information for the IRV method as well as the Borda, Condorcet, and
> Copeland methods. When the data are summarized before submission, the
> individual ranking should be displayed with a means of indicating a tie
> (e.g., font, color,etc).
>
Okay.

> On write-in ballots, how will the voter's input be accepted and matched
> against registered candidates? If there is a perfect match (no spelling
> differences), the ballot can be accepted. But if there is a difference,
the
> possibile matches should be presented (with some identifiers) so that the
> voter can confirm his/her intent. In ranking systems, there is an
incentive
> to submit more than one write-in. Are there limits by state law? We should
> keep it simple, but take account of reality.
>
Here's what I suggest: Let the dummy data include some variations in
spelling, e.g., J. R. Smith, John R. Smith, etc. Our software will cleverly
[*] sort them out and combine them where appropriate (except for a few
remainders). I'd skip the lookup against registered "qualified" write-in
candidates and simply state that some jurisdictions require they be
registered to be counted, and our production software will take all this
into account. I'd also forget about multiple write-ins for a race.

Doug may have some specific useful information on this. I would also run it
by Warren Slocum and other elections people.

We should demonstrate the ability to tally write-in names but also show them
all lumped together as "Write-in."

> The fairvote folks state that you cannot hurt the chances that your
favorite
> choice will win if you rank all of the choices. This is nonsense, but they
> believe it.
>
We could have a dummy dataset where different winners appear with different
scoring methods.

> What we demonstrate about alternative scoring methods should be aimed at
> showing that our intent is to be neutral about voting method religion.
Your
> suggestion that blank rankings be categorized as "no preference indicated"
> is probably better than allowing incomplete ballots--though I see
> "undervotes" as an expression of preference or confusion rather than
error.
> But if we use "no preference indicated," we should allow the voter to
revise
> the input when s/he views the summary before submitting the data.
>
Okay. I agree we don't want to come across as advocating this scoring
method or the other. That's why we show ranked preference in a county race
and not, say, the national race. As far as I know, these different scoring
methods only appear in a few cities and counties. In these cases, we are
offering a practical solution -- not advocating more widespread use. San
Francisco has approved IRV voting but I don't think they've figured out how
to implement it yet.

The fairvote.org people will like us because our system will make it cheap
and easy to implement IRV or similar methods.

> Although I agree that it is easy to overdo item 11, on-screen beauty, I
> think that item should be upgraded from "highly desirable" to "must have."
> I would advise setting a high minimum standard rather than downgrading the
> importance of the item. After all, whatever we can do to make a good
> impression will be appreciated by sophisticated and general
> audiences--including pictures that find their way into online, TV, and
> printed media.
>
Okay. We might want to get in touch with Bob Zeni. He is a graphic artist
that got worked up about the election mess and started something called,
"Design for Democracy."

http://www.electiondesign.org

http://www.aiga.org/resources/content/9/7/8/documents/zeni.pdf

I talked to him once and he seemed agreeable enough.

> We need to work out a timetable for the ballot and reporting work. How
many
> voters should we include in a precinct and how many precincts are we
using?
> Is the central unit a county or a city?
>
I could dummy up some data for you. Say, 500,000 ballots for a county with
1,000 precincts (average about 500 ballots per precinct).

http://home.earthlink.net/~adechert/ballot-mockup3.gif

Our mock ballot would involve county-wide tabulation except for the State
Senate, State Assembly, and U.S. House races. For these three, we can
arbitrarily make up some districts (lists of applicable precincts).

--Alan Dechert
[*] by whatever fakery

==================================================================
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
==================================================================
Received on Sun Aug 31 23:17:16 2003

This archive was generated by hypermail 2.1.8 : Sun Aug 31 2003 - 23:17:18 CDT